Trade policy differentiation between developing countries under GSP schemes

No. 685 – 2005

Norsk

Utenrikspolitisk Norwegian Institute

of International

Arne Melchior

[ 685 ^] Paper

Trade policy differentiation between developing countries

under GSP schemes

Besøksadresse:

Addresse:

Internett:

E-post:

Fax:

Tel:

Alle synspunkter står for forfatternes regning. De må ikke tolkes som uttrykk for oppfatninger som kan tilleg- ges Norsk Utenrikspolitisk Institutt. Artiklene kan ikke reproduseres - helt eller delvis - ved trykking, fotokopi- ering eller på annen måte uten tillatelse fra forfatterne.

Any views expressed in this publication are those of the author. They should not be interpreted as reflecting the views of the Norwegian Institute of International Affairs.

The text may not be printed in part or in full without the permission of the author.

C.J. Hambrosplass 2d Postboks 8159 Dep.

0033 Oslo www.nupi.no [email protected] [+ 47] 22 36 21 82 [+ 47] 22 99 40 00

[Abstracts]

The GSP project

This is a paper from a project on the Norwegian GSP system, undertaken by the Norwegian Insti- tute of International Affairs for the Ministry of Foreign Affairs in 2004-2005. In addition, the follow- ing papers are available:

* Melchior, A., The future of Norway’s GSP system (main report from the project, NUPI Paper No.

680b or Norwegian version in NUPI Paper No. 680a).

* Maurseth, P.B., Trade and development – a selective review, NUPI Paper No. 681.

* Maurseth, P.B., Norway’s trade with developing countries, NUPI Paper No. 682.

* Melchior, A., GSP in the “spaghetti bowl” of trade preferences, NUPI Paper No. 683.

* Melchior, A., The Norwegian import regime for agriculture, NUPI Paper No. 684.

Contact details, NUPI:

* Arne Melchior, +47-99791209, +47-22994038, [email protected]

* Per Botolf Maurseth, +47-48068245, +47-22994040, [email protected]

Arne Melchior

between developing countries

under GSP schemes

1. Introduction¹

GSP is intended to be for developing countries. There is, however, no unambiguously agreed definition of a developing country. When GSP was initiated in the 1960s, the rich man’s club was relatively distinct, and concepts such as “Third World” and “developing countries” were intuitively clear. This clarity is lost for three reasons:

- First, fast growth in East Asia has made formerly poor countries such as Hong Kong, Singapore and Korea climb high on the income scorecards.

- Second, the fall of communism in Eastern Europe has introduced a new group of low and middle income countries in the world economy.

For example, seven of the ten new EU member states had BNI per capita lower than Mexico in 2002, and former Soviet republics are further down on the list.

- A third change is the integration into the world economy of a number of small countries and territories, for example small island states that vary considerably in terms of development.

Facing these developments, a number of new questions arise, for example:

- Are Russia and other former Soviet republics developing countries that should benefit from SDT (Special and Differential Treatment in the WTO) and GSP (the Generalized System of Preferences), immediately or when they become WTO members?

- Should countries that grow richer, such as the Asian examples mentioned above, be “graduated” from GSP and SDT?

- How should small island states with a relatively high income level, such as e.g. Barbados and Antigua, be treated?

The issue is conceptual, technical as well as political.

As shown in Melchior (2005a), there is de facto a considerable extent of differentiation within GSP systems. The types of such differentiation are (i) special treatment of LDCs; (ii) criteria for country or product graduation (in e.g. the U.S. and EU GSP systems); (iii) additional or reduced benefits depending on the adherence to certain standards (e.g.

environmental and social standards in the EU; another list in the USA);

and (iv) what we may call “grey area differentiation”, i.e. extended benefits to specific country groups beyond the LDCs. Examples of this are EU’s preferences for the ACP (Africa, Caribbean, Pacific) countries, USA’s special preferences for Africa under the African Growth and Opportunity Act (AGOA), or Norway’s special preferences for Botswana and Namibia. We call these “grey area” since some of these measures have been subject to specific waivers in the WTO (EU’s ACP regime), and some have been subject to disputes in the WTO (the previous

“countries fighting drugs” scheme of the EU in particular – see discussion below). In this paper, we review some institutional, legal and economic issues related to differentiation. In order to limit the length of the paper, we do not – save for some comments at the end – discuss trade policy differentiation based on governance criteria.

1 I thank Hege Medin and Per Botolf Maurseth for useful comments to an earlier draft.

The standard disclaimer applies.

In the WTO, it is up to countries themselves to “declare” that they are developing countries. In spite of the missing definition of what is a developing country, the concept plays an important role in the WTO legal system, with a number of special provisions. Beyond the general concept of developing countries, the only special category accepted is the Least Developed Countries, which benefit from various forms of extended preferential treatment.²

There are nevertheless references to various other forms of differentiation in the WTO legal frameworks; for example:

- In the Appendix of the Uruguay Round (UR) Agreement on Subsidies, countries with BNI per capita below 1000 USD are given special rights.

This suggests that income-based criteria may in some circumstances be legally acceptable in the WTO.

- In the same agreement, developing countries with a world export share below 3.25% for a specific product may be subject to less strict disciplines.

Such market share criteria are extensively used in GSP product graduation. They are however controversial among GSP beneficiaries that do not enjoy losing their market access. According to some “.. sector and country graduation are contrary to the principles of non-graduation and non-reciprocity that underpin the GSP and therefore alien to the original intentions underlying the GSP concept” (see WTO/COMTD/W/77 p. 21).

- In the UR Agreement on Agriculture, “net food-importing developing countries” is a particular category.

This is however linked to this specific agreement, and has not been used for SDT generally.

In the mandate from Doha in 2001, there are various references to sub- categories. For example, the WTO should (our italics added)

- “..frame responses to the trade-related issues identified for the fuller integration of small, vulnerable economies into the multilateral trading system, and not to create a sub-category of WTO Members..”

- There is also a reference to “..steps that might be taken within the mandate and competence of the WTO to enhance the capacity of the multilateral trading system to contribute to a durable solution to the problem of external indebtedness of developing and least-developed countries.”

- On technical assistance, it is stated “The delivery of WTO technical assistance shall be designed to assist developing and least-developed countries and low-income countries in transition to adjust to WTO rules and disciplines, implement obligations and exercise the rights of membership, including drawing on the benefits of an open, rules-based multilateral trading system. Priority shall also be accorded to small,

2 For an overview of legal WTO provisions on SDT and the LDCs, see WTO (2000).

There is a sizeable literature on SDT; see e.g. Whalley (1999), Michalopoulos (2000), Hart and Dymond (2003), Hoekman et al. (2003). For earlier contributions, see e.g.

Balassa (1980) and Whalley (1990).

2. WTO agreements on differentiation

vulnerable, and transition economies, as well as to members and observers without representation in Geneva.”

Hence the Doha mandate mentions smallness, vulnerability and indebtedness, and refers to transition countries. Observe that the WTO ministers explicitly reject the notion of creating a subcategory of small countries in addition to LDCs. In the Doha round, there is however a work programme on small economies.

Hence there are hints about and references to various categorisations, but only LDCs have obtained a special status. Any change in the current rules would imply conflicts. Specifying graduation criteria for the more advanced developing countries trigger opposition from countries that may lose some of their benefits as well as the institutional affiliation to the developing country group. Appointing new categories for special treatment creates distributional conflicts among those that are eligible and those that are not.

GSP is unilateral and non-reciprocal and the donors therefore have greater leeway as to what may be done. As we shall discuss below, there are also limitations as to what kind of discrimination that is legally possible under GSP. With the current practice of numerous “grey area”

measures in the field, an issue is whether conditions for discrimination should be more “streamlined”. This is also a matter of technical trade barriers: If every GSP donor practices discrimination according to varying concepts, for tariffs and possibly also rules of origin, the world trade system will really become a “spaghetti bowl”, as some have called it already. The system would be simpler if multilateral rules on differentiation were made clearer.

As a point of departure for our examination, it is of interest to review some of the different categorisations and definitions related to developing countries that have been applied in various contexts.

The list of LDCs is managed by the UN Economic and Social Council (ECOSOC). The criteria for inclusion are currently³

- 3-year average GNI per capita below 750 USD;

- score below a certain level for a Human Asset Index (HAI) covering nutrition, health, education and adult literacy;

- score above a certain threshold for an economic vulnerability index (EVI), covering fluctuations in agricultural production, fluctuation in exports, smallness, export concentration, the exposure to natural disasters and finally (less clear to us) the economic importance of non- traditional activities;

- countries with a population above 75 million are not eligible.

The LDC list is reviewed tri-annually, and ECOSOC then sets the threshold levels. In order to be included, countries must conform to all three criteria. If an existing LDC fails to pass the test for at least two of the

3 See the web page of the UN Office of the High Representative for the Least Developed Countries, the Landlocked Developing Countries and the Small Island Developing States, http://www.un.org/special-rep/ohrlls/ohrlls/default.htm for more details.

3. Subgroups of developing countries – a smorgasbord

criteria in two consecutive reviews, it would normally be graduated.

Botswana was the first country to graduate in the 1990s. Senegal and Timor-Leste have been added to the list recently, while the Maldives, Cape Verde and possibly (pending the next review) Samoa will graduate.

Among the LDCs, Cape Verde, Samoa and Vanuatu have GNI per capita above 1000 USD, while the Maldives is at 2170 USD. Hence some of the LDCs are lower middle income countries.

Beyond the LDC list, there are currently a number of categorisations used for different purposes of development policy. In the WTO work programme on small economies, the group of landlocked developing countries (LLDCs) have also voiced their interests. These countries coordinated their views in a Ministerial meeting in Almaty, Kazakhstan in 2003. In the Cancun Ministerial meeting in 2003, the Trade Ministers of the landlocked developing countries presented their views. This was followed up in UNCTAD 2004⁴, and the UNCTAD Secretariat has undertaken work in the field (UNCTAD 2003).

The LLDC initiative came as a surprise to the Small Island Developing States (SIDS), which have also been fighting to obtain better market access. The SIDS concept was established by a UN conference at Barbados in 1994, leading to the “Declaration of Barbados” and a Programme of Action for the SIDS (Hein 2004). The UN has appointed its own “High Representative for the Least Developed Countries, the Landlocked Developing Countries and the Small Island Developing States”. At least four lists of SIDS exist (see Hein 2004, Encontre 2004), however, and the merit of SIDS as a particular category is disputed. Some of the UN lists of SIDS actually include countries that are not so small (Haiti and the Dominican Republic have populations at 7.5 and 8.7 millions, respectively), that are not islands (a few continental countries were included), not so developing (e.g. Singapore is on the list), and not states (a number of territories are included) (see Encontre 2004).

Given the increased focus on poverty reduction in global politics, LLDC and SIDS are not the only relevant categorisations beyond LDCs.

One option would be to stick to the World Bank’s income classification, shown in Table 1:

4 See Ministerial Declaration, UNCTAD document TD/409, 13 June 2004.

Table 1: The World Bank income classification Income range

Category

Abbre- viation

used

here 2003 2004

Number of coun-

tries, 2004

Examples of countries included,

2004 Low-

income LIC ≤765$ ≤825$ 59 Bangladesh, India, Pakistan Lower

middle income

LMC 766-

3035$ 826-

3255$ 54 Brazil, China, Egypt, Indonesia, Namibia, Thailand, Ukraine Upper

middle income

UMC 3036- 9385$

3255-

10065$ 40 Argentina, Botswana, Malaysia, Russia High

income none >9385$ >10065$ 55

Bahamas, Hong Kong, Korea,

Sinngapore Source: http://www.worldbank.org/data/countryclass/classgroups.htm

Relating to this classification, we should observe that:

- The income thresholds have to be adjusted each year to account for inflation. Inflation adjustment is made using average inflation in G-5 (the “SDR deflator”).

- Second, exchange rate changes may lead to considerable fluctuations for individual countries. The World Bank uses a specific methodology (the “Atlas method”) to smooth out such fluctuations, but they still matter. For example, Korea has an income according to 2003 data at 12030 $. In 2002, however, it was only 9930$. The strong change was partly due to the dollar depreciation, and it illustrates the role of annual exchange rate changes .

A more fundamental question is where the income thresholds come from.

Why is the low-income threshold now 825$ and not, say, 900$?

According to the World Bank, its thresholds were mainly set 30 years ago as lending criteria. The low-income category is the threshold for allowing domestic preferences in bidding under World Bank lending.⁵ The threshold between lower and upper middle income countries is the threshold for 15-year IBRD loans. The high-income threshold was set at 6000$ per capita in 1989. The original thresholds were based on a study of the relationship between income and other indicators of well-being, such as poverty incidence and infant mortality.⁶ The choice of thresholds was also influenced by “resource availability”. We interpret this in the sense that the low-end group should not be so broadly defined that it would be a problem for lending capacity. There has been some later adjustment, but mainly just to account for inflation.

Hence even with the World Bank income criteria, we cannot escape from the fact that thresholds have to be based on some other considerations than income. Table 2 shows some other “hybrid” country

5 See §2.55 in World Bank, 2004, Procurement under IBRD loans and IDA credits, available at www.worldbank.org.

6 See http://www.worldbank.org/data/countryclass/history.htm.

classifications used for aid purposes, based on income combined with other criteria:

Table 2: Other classifications related to developing countries

Name Criteria/ explanation Number

LICUS:Low- income

countries under stress

Low-income countries with particularly weak governance, with a low score on certain indexes used by the World Bank.⁷ Used by the World Bank recently.

13 (2003)

IDA

(International Development Association)

This is the list of countries qualified for interest-free loans and grants from, an affiliate of the World Bank.

There are currently 81 IDA countries.⁸ IDA countries are selected on the basis of income (currently BNI per capita below 865 USD), creditworthiness and indicators on governance.

81

Severely indebted countries

Regular lists published by the World Bank. For a country to qualify, the present value of debt service to GNI must exceed 80%, and the present value of debt service to exports must exceed 220%.

52

HIPC:

Severely

indebted poor countries

Among the 52 severely indebted countries (list published annually by the World Bank), 42 are classified as HIPC. This is related to the debt relief measures under the HIPC programme of the World Bank, where countries also engage in Poverty Reduction Strategies – based on Poverty Reduction Strategy Papers (PRSP). To become a HIPC country, a country must qualify under IDA (see above) or IMFs Poverty Reduction and Growth Facility, and the country must have an unsustainable debt situation after standard debt relief measures under the Paris Club have been attempted.

42

ODA (Official Development Assistance)

For statistical purposes, the Development Assistance Committee of the OECD (DAC) publishes a list of aid recipients, with two different categories. Part I is ODA (Official Development Assistance) and Part II is Official Aid (OA).⁹ The latter includes some transition countries (Russia, Ukraine, Belarus, and in 2003 also some of the forthcoming EU members), and a category of 25 “Advanced Developing Countries and Territories” including e.g. Singapore, Hong Kong, Korea and Israel. There is only one high- income country (Bahrain) on the ODA list, hence income per capita is an important criterion.

147

7 They score below 3.0 for the Country Policy and Institutional Performance Assessment

indicator, see http://web.worldbank.org/WBSITE/EXTERNAL/PROJECTS/STRATEGIES/EXTLICU

S/0,,contentMDK:20176979~menuPK:511786~pagePK:64171531~piPK:64171507~theS itePK:511778,00.html#status_determined. On the other hand, there are not unambiguous thresholds for obtaining LICUS status.

8 About criteria for eligibility, see

http://web.worldbank.org/WBSITE/EXTERNAL/EXTABOUTUS/IDA/0,,contentMDK:

20189587~menuPK:413944~pagePK:83988~piPK:84004~theSitePK:73154,00.html

9 See www.oecd.org/dac/stats/daclist

ODA is mainly a developing country definition that excludes high- income countries and some transition countries. IDA is a category that overlaps to a considerable extent with the low-income category of the World Bank. Most low-income countries are included in IDA, and in addition around 20 others – including Indonesia and some Eastern European countries. HIPC and highly indebted countries are not “general”

classifications since they focus on debt, but in the context of trade they could be used as supplementary criteria if there is a wish to take into account indebtedness.

Hence with LDC, LLDC, SIDS, LIC, LMC, UMC, LICUS, HIPC, IDA and ODA there is an array of different categories that might potentially be used for discriminating between developing countries. In addition, one may use technical criteria, such as e.g. thresholds for market shares in the case of product graduation. Or one may use directly indexes for governance, indexes for vulnerability etc. that are used for constructing the various country lists. We shall revert to the issue of how plausible these criteria are.

Say, hypothetically, that Norway would choose to give extended tariff preferences to any of these groups under GSP. Would this be allowed according to the WTO? In the following, we review some recent legal developments in the WTO that is relevant for the assessment.

4. Legal background: The India-EU dispute on GSP

GSP was allowed by a waiver in GATT until 1979, when the enabling clause was decided in the Tokyo Round. The enabling clause is not part of the GATT Agreement as such, it is a so-called decision (see GATT 1980, 203). The enabling clause also refers to the GATT decision of 1971 on the establishment of “generalized, non-reciprocal and non-discriminatory”

preferences”. The terminology is important: While “generalized” is sometimes erroneously spelled as “general”, the term – together with the word non-discriminatory – refers to preferences that apply to developing countries in general and not to specific groups (for example former colonies). The right to such preferential treatment applies to tariffs under GSP, and to non-tariff measures governed by multilaterally negotiated GATT provisions.¹⁰

After the GSP systems were established in 1971, EEC introduced the Lomé convention for African, Caribbean and Pacific countries. Since the ACP preferences are not “generalized”, the EU needed a special waiver for this. This waiver has been extended; last time under the Doha Ministerial meeting in 2001. The ACP trade regime, currently under the Cotonou Agreement, was then granted a waiver until end-2007. Other WTO members only accepted this waiver after the EU agreed to modify its import regime for bananas, to the advantage of non-ACP suppliers.¹¹

10 In the light of the current “preference erosion” debate, it is of interest to observe that paragraph 3b of the enabling clause states that SDT “shall not constitute an impediment to the reduction or elimination of tariffs or other restrictions to trade on a most-favoured nation basis” (ibid., 204).

11 See www.wto.org, WTO document WT/MIN(01)/15, European Communities — the ACP-EC Partnership Agreement, decision of 14 November 2001.

The EU is currently negotiating so-called Economic Partnership Agreements (EPAs) with the ACP countries. These will have a stronger element of reciprocity, and their WTO-legality may then be secured through WTOs rules for free trade agreements.

In addition to the “everything but arms” (EBA) regime for the LDCs and the ACP regime, the EU has also given extended trade preferences for two other country groups:

- The Overseas Countries and Territories (OCT), consisting of 21 areas with small populations.

- The “Countries fighting drugs” group, comprising 11 Latin American countries and Pakistan.¹²

In 2002, India challenged the EU “drugs” regime for being discriminatory (see textbox).

India also challenged discrimination based on labour rights and environmental standards, but later withdrew the complaint.¹³

After panel ruling and appeal by the EU, the WTO Appelate Body finally decided that the EU drugs regime was inconsistent with WTO rules. This ruling is important because it clarifies to some extent the legal options for GSP. Our interpretation of the Appellate Body (WTO 2004) decision is as follows:

- A GSP donor may give different treatment to different developing countries.

- Such discrimination without any explicit

authorisation is possible only for the LDCs.

- For other developing countries, discrimination has to be based on objective criteria, so that all developing countries fulfilling these criteria obtain similar benefits.

- The criteria have to be related to the “development, financial and trade needs” of developing countries. Here the Appellate Body overturned the earlier Panel decision: According to the Panel, the needs of all

12 On OCT, see Official Journal of the European Communities (OJ) 30 November 2001, on Drugs, see OJ 31 December 2001 and various information material, e.g. EU Commission (2004).

13 The reason was apparently that India itself had not been “materially affected” by this;

not that India “gave up” the case.

India’s complaint about EU’s GSP In its complaint (WT/DS246/1, G/L/521, 12 March 2002), India e.g. stated:

“India is particularly concerned about the following conditions:

1. The tariff preferences accorded under the special tariff arrangements for combating drug production and trafficking are available only to specified countries selected by the EC; and

2. The tariff preferences accorded under the special incentive arrangements for the protection of labour rights and the environment are accorded only to countries that meet labour and environmental policy standards determined by the EC.

India considers that the tariff preferences accorded under these special arrangements create undue difficulties for India's exports to the EC, including for those under the general arrangements of the EC's GSP scheme, and nullify or impair the benefits accruing to India under the most-favoured-nation provisions of Article I:1 of the GATT 1994 and paragraphs 2(a), 3(a) and 3(c) of the Enabling Clause.”

developing countries had to be considered; the Appellate Body decided that the consideration might apply to less than all developing countries (ibid., 64, 71).

- The objective standards cannot be any assertion about needs, but

“Broad-based recognition of a particular need, set out in the WTO agreement or in multilateral instruments adopted by multilateral organizations, could serve as such a standard.” (ibid., 66)

- Finally, such discrimination should not impose unjustifiable burdens on other Members.

EUs drug scheme was inconsistent with the enabling clause because the benefits were not available to all developing countries facing a similar situation. If the EU had created a drug regime with clear rules for inclusion and exclusion, available to all developing countries, it is likely that it would have passed the legal test. The EU has chosen not to do so; it abandoned the drug regime from 1 July 2005 as part of the current revision of its GSP system.¹⁴

Which of the smorgasbord presented in the last paragraph would be acceptable in the light of this legal interpretation? If e.g. Norway granted extended preferences to LLDC, SIDS, LIC, LICUS, HIC, HIPC, LMC, IDA or ODA; which list would be WTO-legal? It is impossible to tell this with certainty, since WTO law interpretation may be modified and developed in later disputes, or by WTO decisions. Most of these lists are based on objective criteria, and if these criteria were made explicit in cases where this openness is missing, there should be no problem arguing for the development relevance of the criteria involved. The LLDC and SIDS definitions might have to be sharpened for such a purpose.

Some of the country groups referred to here were constructed for non- trade purposes, and some of the more narrow cases (e.g. LICUS, HIPC) do not seem fit to be major criteria in a trade context. Hence if such classifications were to be used in the GSP context, the pure income criteria (LIC, LMC) or income with supplementary criteria (IDA, ODA) may be relevant. It would also be possible to construct any combination of criteria (for example, IDA+LIC+LLDC, or the like).

5. Characterising groups

Having established that some of the categorisations mentioned above might be WTO-legal, we approach the issue of whether discrimination of any kind is actually warranted. As a first step, we shall study the differences between the various lists. Table 3 gives more intuition about the characteristics and relative importance of the different groups. In the table, the income-based definitions use 2002 data for GNI per capita, from the World Bank (Atlas method). The categories in the table are:

- LDC, LIC, IDA: As explained above. LDC and IDA based on current lists.

- 1000$: All countries with GNI per capita below 1000 $.

- LIC+LMC: Low income + lower middle income countries.

14 See www.wto.org for material on dispute No. DS246 for documentation on deadlines, arbitration etc., and Official Journal of the European Communities 30.6.2005 on the new GSP regime of the EU.

- +UMC: LIC+LMC+UMC: Also including upper middle income countries.

- +HIGHINC: LIC+LMC+UMC+Korea+non-OECD high-income countries.

These seven categories are more or less alternative definitions of developing countries, ranging from extremely narrow (LDCs) to very broad (including all high-income countries. In the table, we also include the “supplementary” categories SIDS, LLDC and severely indebted. These are:

- SIDS 1: SIDS definition with 37 countries from the web page of the UN High Representative.

- SIDS 2: Unofficial UNCTAD list of 29 countries from Encontre (2004).

- LLDC: List of 30 countries from UNCTAD.

- Severely indebted: 52 countries based on current World Bank classification (May 2005).

In the table, it should be observed that the number of observations varies between columns. Only for population do we have 100% data coverage.

The less data, the less accurate are the results; hence it should be observed that the shares are based on the available data only. In general, however, we believe that the table provides a reliable picture, and the data coverage is rather high. In the table, we show the percentage of world population covered by data on each variable. This is between 90 and 100% for all variables except poverty. For poverty, the figure is 77-78%. Here it should be recalled that poverty data are not systematically provided for rich countries, which account for 15% of the world’s population. Given that there are few people living below 1$ a day in rich countries, the “real”

data coverage is certain to be above 90%.

Table 3: The relative size of various categories of developing countries:

Share of world total (%) for different variables.

Category Imports Exports Agric.

Exports

Income (GNI)

Income (PPP)

Popu- lation

Poor 1$

Poor 2$

LDC (50) 0.71 0.60 0.67 0.53 1.76 11.09 17.04 12.90

LIC (61) 2.48 2.35 2.86 2.74 9.26 36.63 68.64 59.24

IDA (81) 3.34 3.41 4.69 3.49 11.03 40.80 70.71 64.28

1000$ (67) 7.80 8.59 8.15 7.48 23.17 60.38 91.64 88.49

LIC+LMC (117) 15.44 16.58 20.01 13.70 36.85 79.14 98.16 98.15 +UMC (154) 23.69 25.67 29.96 18.91 43.44 84.47 100.00 100.00 +HIGHINC. (182) 34.14 37.12 31.68 21.67 47.42 86.04 100.00 100.00

SIDS (1) (37) 2.27 2.23 1.31 0.46 0.59 0.83 0.02 0.04

SIDS (2) (29) 0.27 0.17 0.35 0.11 0.15 0.26 0.02 0.03

LLDC (30) 0.59 0.53 1.16 0.33 1.12 5.44 10.31 7.21

Severely indebted (52) 3.67 4.06 10.81 3.75 7.89 14.37 17.04 9.56

Observations 216 216 150 172 229 233 97 88

% of world pop.

covered by data 100 100 91 97 100 100 78 77

Year of data 2002 2002 2002 2002 2002 2002 2001+other

Data source WDI WDI COM-

TRADE WDI WDI,

CWF WDI WDI Table 5 WB web pages Note: WDI= World Development Indicators 2004, or later WDI data on the World Bank’s web pages (for poverty). CWF=CIA World Factbook (also used to supplement trade and income data).

The LDCs is a small group in economic terms, covering 0.5-0.7% of world trade or nominal income, 1.8% of real income, 11% of the world’s population, and 13-17% of the world’s poor. With the reservation that LDC may be slightly underrepresented due to missing data, only a modest proportion of the world’s poor are in the LDCs. Our poverty headcount data at the country level cover 986 million people at 1$ a day, and 168 of these are found in the LDCs.¹⁵

In low-income category, the populous countries in South Asia are added (in addition to Bangladesh, which is an LDC), and all shares are approximately multiplied by four. The LICs include 2/3 of the world’s poorest, but less than 3% of the world economy in nominal terms.

With IDA, we add Indonesia and 20 other countries, but the shares are only modestly affected. Moving to 1000$, however, China is included.

The figures on economic size are then doubled, and another 1/5 of the world’s poor is added. Including the whole group of lower middle income countries, including Brazil, Russia, Thailand and others, the economic importance increases further to 14-20%. Now we include almost all global poverty.

15 For the LDCs, we have poverty data for countries representing 2/3 of their combined population. If the poverty incidence is the same for the countries with missing data, the figure would increase to 254 million. With complete data for all countries, it is likely that the share of poverty in the LDCs would be above 1/5. Even if this is a substantial share, it remains true the majority of poor people worldwide are outside the LDCs.

Adding the upper middle income countries – including countries such as Argentina, Malaysia and Mexico – increases the economic size further, but there are just a few more poor people to include. The high-income countries that are normally considered as developing countries in the WTO constitute a relatively small group in economic terms, even if Korea is included in our data. With the broadest possible definition of developing countries, including Korea and high-income countries outside the OECD, the developing countries cover approximately 1/3 of world trade, and 22%

of world nominal income.

Most people would agree that defining developing countries (DCs) as the LDCs would be too narrow, and that including high-income countries in the DC concept may be questioned. There is however no scientific solution to this problem; any threshold is adjustable and depends on the criteria we set. For example, Norway has BNI per capita almost three times higher than Korea. So Korea may legitimately feel “DC” compared to Norway.

Using such income thresholds, the data also illustrate that some countries switch position each year. For example, Russia was lower middle until 2004, when is moved to upper middle. The same occurred for Botswana, that was even an LDC until the early 1990s.

In research on the world income distribution, a well-known feature is the existence of a “bimodal” distribution with few in the middle (see e.g.

Quah 1996). If we undertake a “cluster analysis” to see “which countries belong together”, is likely that a cut-off point between North and South will appear somewhere in the middle. Diagram 1 illustrates this “sparsely populated middle”, showing the cumulative share of world exports, with countries ranked by income levels. The diagram only shows the lower range of the distribution (142 out of 171 countries where we have data on both variables).

The diagram is also relevant in the context of trade policy, by showing

“how much competition” that will come if trade preferences below certain income thresholds are granted.

Diagram 1: The poor countries' share of world exports (2002) (142 countries and territories, ranked by BNI per capita)

Data sources: World Development Indicators 2004, CIA World Factbook.

0 5 10 15 20 25 30

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 GNI per capita (USD, Atlas Method, 2002)

Cumulative share (%)

Indonesia (710) China (960)

India (470)

Thailand (2000) Russia (2130)

Brazil (2830) Botswana (3010)

Malaysia (3540)

Mexico (5920)

Korea (9930)

Namibia (1790)

In the income ranges up to 4500 USD, there are many countries (127 for which we have data). After that, however, there are fewer. Between 4500 and 10000$ there are only 15 countries. Hence the diagram illustrates the

“sparsely populated middle” in the world income distribution. Above 10000$, we find 29 rich countries (not shown in the diagram) that represent more than 70% of world exports.

In the Diagram, we also show the positions of Botswana and Namibia, that obtain special preferences in the Norwegian GSP system. Both were LMC based on 2002 data, but – as noted – Botswana has now moved to UMC. Botswana is richer than Brazil and a number of other large agricultural exporters. According to “objective criteria”, it may be difficult to explain this differential treatment.

A diagram showing shares of world income would be quite similar.

For shares of world agricultural exports, the curve is also similar, but with some modifications for the DCs, shown in Diagram 2.

Diagram 2: The DC's share of world agricultural exports, 2002 Based on data for 133 countries (104 shown)

Data sources: World Development Indicators 2003, COMTRADE.

0 5 10 15 20 25 30 35

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 BNI per capita (current USD, Atlas method, 2002)

Cumulative share of world agricultural exports (%)

China

Thailand Brazil

India Indonesia Turkey (2490)

Malaysia Argentina (4220)

Mexico

Agriculture is important for more countries so that some of the flat segments in Diagram 1 disappear (e.g. in the income range between Indonesia and China). The relative importance of the large countries is also changed; Brazil is more important than China for agriculture, and Argentina (GNI per capita at 4220) is a major exporter (the third largest after Brazil and China). Third, developing countries with an income level above 6000$ do not add much to the DC share: Beyond 6000$, the curve is flat.

The diagrams shed light on Table 1 above. The economic size of each group is strongly affected by which larger countries that are included. For example, the LIC group is much larger than LDCs mainly because it includes India and Indonesia. With the 1000$ threshold we include China, and the economic magnitudes jumps sharply. In the lower middle income group we find Thailand, Russia and Brazil, which also add significantly to the cumulative share of LIC+LMC. In the +UMC group we add 37 new countries including e.g. Malaysia, and Mexico, and Argentina for agriculture, so the trade share rises further.¹⁶

Diagrams 1 and 2 illustrate that if we use statistical techniques to identify clusters, we are likely to find a cut-off point between rich and poor in the 5-10000$ range. This cut-off point is likely to be inside the

16 Table 1 does not show ODA, but this would be close to the +UMC category (somewhat below due to the exclusion of Russia etc.).

upper middle income category.¹⁷ The ODA classification conforms with this, except that Russia and many other Central/Eastern European countries statistically belongs to the lower group.

6. Reasons and criteria for differentiation

The basic idea behind unilateral preferences is that market access is s good thing that should be given to those that need it more. The definition of who needs it more may be based purely distributional reasoning (the poorest should have more), or “functional” reasoning – e.g. that countries should be compensated for particular problems. In the functional reasoning we find concepts such as smallness, landlocked-ness, indebtedness, remoteness and vulnerability. As seen above, vulnerability is defined by various characteristics such as exposure to natural disasters, income volatility etc. The LDC vulnerability index in fact also includes smallness.

In terms of economic theory, the concepts of smallness, landlocked- ness and remoteness may all be defended in the light of the new trade theory and economic geography. For example, Redding and Venables (2004) show that being landlocked reduces income by 24%, being an island by 7% and closing the borders by 20-27%. Being remote also matters: Moving Zimbabwe to the location of Hungary would increase its income by 80%. All these effects are due to various forms of scale economies. Being economically remote is however only partly determined by nature; if all African countries grow economically, Zimbabwe will no longer be so remote. Remote Pacific islands might however have a problem of remoteness by nature; there is not much land around. Land transports are on the whole more costly than sea transports (WTO 2002), and this may explain why landlocked countries are worse off than islands.

In terms of transport costs, islands are nevertheless worse off than continental countries with access to the sea (ibid.). The impact of smallness is more ambiguous; Easterly and Kraay (1999) e.g. find that micro-states experience more volatility but are actually better off economically, on average. WTO (2002) also finds a similar relationship.

facing this, it is no surprise that the SIDS proponents have had problems to convince the world community about the merits of their message.

Encontre (1999) also notes, in his analysis of SIDS, that the frequency of natural disasters did not have a statistically significant impact on growth.

In his defence for the SIDS concept, he focuses on the combination of smallness, remoteness, high transport costs, a narrow resource base and non-diversified economies.

Even if some of the criteria for special treatment are supported, it is not certain that trade policy or the GSP system is the right way of addressing these problems. If poverty is the problem; one should not need criteria related to the causes as well. Hence if e.g. smallness leads to

17 From non-Norwegian sources (and not the OECD itself), we have had access to the OECD Secretariat cluster analyses undertaken in 2001-2002 in order to examine criteria for trade policy graduation. These created considerable controversies since OECD includes Korea, an obvious candidate for graduation. The discussion above illustrates why cluster analysis is likely to put Korea in the rich group.

poverty in some cases, it is the effect that matters. Hence even if some the functional arguments are sensible, it is not so clear how much can be won by adding these criteria in addition to income. It is true that income is not a perfect measure of well-being, but it is a simple indicator that is correlated with a large number of other “good things”: Easterly (1999) examines 81 alternative indicators of living conditions, and finds that 61 of these were positively and significantly correlated with income, and in 32 case income was a strong determining factor. Hence in spite of some weaknesses, income is a universal criterion that is simple. By using income, one also avoids that all WTO members come with their alleged handicaps, in order to obtain better market access. The current WTO policy of “no more special categories” therefore seems to be well advised.

Even if we conclude, “income is the best proxy we have”, the problem remains that the income thresholds do not follow. As noted above, the World Bank income thresholds were originally based on additional criteria related to living standards. Hence the World Bank income categories are not clean measures of income; the thresholds are affected by other criteria.

These thresholds are created for specific purposes, and it is not evident that the low-income level of the World Bank is better than e.g. 1000$.

Hence we should treat all thresholds with some pragmatism. Since these thresholds are in some sense arbitrary, it would be an advantage to design trade preferences so that the difference due to such thresholds is too dramatic. For example, if tariffs are cut from 15 % to zero for LDC+LIC+IDA, they could be cut to 5% for LMC, and to 10% for UMC.

In this way, the impact of thresholds is not too dramatic, and measurement errors in GNI per capita will not have too dramatic effects.

If income thresholds are used pragmatically in this way, an issue is whether they should be supplemented by other measures. Concerns for financial aspects is written into the WTO framework, and reflected in e.g.

IDA or HIPC. Along with income, this is also a measure of economic performance rather than a cause. “Financial needs” is already written into the enabling clause, and there seems to be no legal or conceptual problem of adding debt criteria. Using IDA, the may not be a need for this.

If we rely on income as a main criterion for differentiation, we are back to the distributional reasons for differentiation. A fair criterion would then be to have GSP benefits that are inversely related to poverty. Given that it is impractical to have one tariff for each country, a possible option is to differentiate by classes, e.g. (i) LDC, (ii) LIC+IDA, (iii) LIC+IDA+LMC, (iv) +UMC. How many steps is a pragmatic consideration; with 5% MFN tariffs it is better to give zero tariffs to all.

But in a situation with gradual liberalisation, such an approach might be taken.

Instead of this variety of “fairness”, another approach would be the

“Rawlsian” one – to give it all to the poorest. Having concern only for the LDCs would be the trade policy equivalent. Considering the amount of poverty outside the LDCs, most people would find this to be an implausible argument. As stated by Hoekman et al. (2003), “ preferences should focus on the poor, wherever they are located, and not on a limited set of countries”.

7. The political economy of “supply constraints”

When the World Bank set its income thresholds, resource availability was also an aspect considered. For trade policy, a similar concern for import competition or the “cost of liberalisation” may apply. For example, countries may fear import competition from India, Brazil or China, and therefore prefer to give generous trade preferences only the LDCs. If there is a political pressure to help the poor, a focus on the LDCs may then act as a lightning rod. In the debate on agricultural protection in Norway, some people e.g. argue that Norway should reduce trade barriers in order to help developing countries. According to the farmers’ lobby, however, Brazil is not a “real” developing country, and we should focus on the very poorest countries. Facing competition from the “industrial” agriculture of Brazil, small-scale farmers in the LDCs should form an alliance with farmers in Norway in order to avoid too much liberalisation. While we do not question the true concerns about poverty expressed by Norwegian farmers, we believe that it is too narrow to focus only on the LDC’s, that account for 0.6% of world exports or a meagre 0.2% of Norway’s imports.

In the following, we shall use an international trade model in order to shed some light in these issues. Experience with trade preferences to the poorest suggests that the trade created is modest, and this is in turn interpreted as an indication of “supply constraints”. As a theoretical illustration of supply constraints, we may use a model of Melchior (2004):

This is a neoclassical trade model where two countries use capital and labour in the production of two different goods; one labour-intensive and the other capital-intensive. In the model, the poor country may become fully specialised in the L-intensive good if is has too little capital.

Furthermore, if its capital-labour ratio is below what is most efficient in the production of L-intensive goods, the poor country becomes a less efficient producer. Hence the supply limitation of the poorest country is created by a too low capital-labour ratio. Since the poorest countries have too little human or physical capital, they are less efficient and export less to the North. As the capital-labour ratio in the South grows, it becomes more efficient and labour-intensive exports grow. At some stage, however, the country starts producing the capital-intensive good as well, and exports of L-intensive goods decline as the K/L ratio increases further. In this range, factor prices become equalised if there is free trade.

Diagram 3 shows, by numerical simulation of the model, how exports of L-intensive goods from the South to the North depend on the capital- labour ratio in the South, for a given capital-labour ratio in the North (equal to k2 in the diagram).

Diagram 3: Exports of L-intensive goods from South to North

0 10 20 30 40 50 60 70 80 90 100

0.0 0.1 0.2 0.3 0.4

Capital-labour ratio Exports of L-intensive good from country 1

South specialised in L-intensive good

Factor price equalisation

k_1A k₂

China

Korea LDCs

Here the horizontal axis measures the capital-labour ratio in the South, k2

is the similar ratio in the North, and k1A is the turning point when the South starts producing both labour- and capital-intensive goods. Diagram 4 shows the corresponding wages in the North and in the South.

Diagram 4: Wages in the North and the South

0 0.2 0.4 0.6 0.8 1 1.2

Capital/labour ratio in the South

Wage level

Wage in the North

Wage in the South

South specialised in L-intensive good

To the left in the diagrams, where the South is inefficient due to “too little capital”, imports is not much of a threat in the North, but with a rising capital/labour ratio, the South becomes more efficient and this reduces the Northern wage. To the right of the horizontal line when the South

becomes diversified, the two wage curves coincide since there is factor price equalisation.

Although the model is made with two countries only, we may use it to shed some light on “supply capacity” and its possible impact on trade policy. We may think of LDCs as located far to the left in Diagram 3, China somewhere along the rising portion of the curve, and Korea on the falling curve to the right. Observe that when the “threat of imports” from China is at its largest, further capital accumulation in China will lead to a reduction in L-intensive exports. Hence when the threat of import competition is at its greatest, it disappears. In a sense, the best way of getting rid of competition from China is to let it grow further and accumulate more capital, so that its wage level rises. Due to the size of China, this may take more time than it did in e.g. Hong Kong or Korea, but the principle should be the same. Some analysts have predicted that China could experience labour shortage in 10-15 years.

For LDCs, the presence of large labour-abundant competitors like China leads to a downward pressure on wages and world prices of labour- intensive goods. The best way of changing this situation will also be to let China grow further so that its wage level rises. In order to help the LDCs, we should let China follow Korea to the right in the diagram.¹⁸ According to this line of reasoning, it would be a bad idea to have a trade policy that prevents further growth in China, by restricting imports from China. If LDC trade is allowed while China trade is restricted, the static gain for LDCs might be larger, but in the long run, the LDCs will gain from fast structural change and rising wages in China.

In the case illustrated above, the LDC has the most limited supply capacity, and improved market access for this will generate less North- South trade than free trade with China. According to the model, it can be shown that the relative welfare improvement will be larger for the LDC than for China. In the North, welfare gains are greater in the case of trade with China, but the impact on wages is stronger.¹⁹ If welfare considerations determine trade policy in the north, it should allow imports from China at least as much as imports from LDCs. But if labour in the North influences trade policy, it would prefer to trade more with the LDC since it makes less harm. Hence the model is a simple illustration of the political economy involved: Workers, or farmers, may prefer free trade with LDCs because they fear the competition from China or Brazil.

According to Hoekman et al. (2003), it is economically sensible but politically unrealistic that rich countries will provide unlimited preferential market access to large countries such as China and India. For this reason, they suggest a greater focus on multilateral liberalisation for goods of particular interest to developing countries in general. This illustrates the interaction between trade preferences and multilateral trade rules. On the other hand, precisely this interaction has also resulted in the conflict on

“preference erosion” – where countries with favourable trade preferences resist multilateral trade liberalisation that will undermine these advantages.

18 This does not follow from the model, which contains no dynamics and only two countries; we only use the model as a device to structure the discussion.

19 For details, see article on www.nupi.no.

8. Preference erosion

As a consequence of this potential alliance between LDCs and workers in the North, there is even a risk that trade policy could become blocked in a situation with free trade for the poorest only, and more restricted trade for the more efficient developing countries. The current debate on preference erosion; where some of the poorest WTO members, is an indication of this problem. Although we agree that the problem of preference erosion is real for some poor countries and should be addressed, it seems wise to avoid such problems in the future.²⁰

The “preference erosion” debate is particularly related to EU preferences for the ACP countries. In the Uruguay Round of trade negotiations, some Caribbean countries feared that multilateral trade liberalisation could undermine their preferences. In the current “Doha development agenda”, similar concerns have been voiced. Table A in the Appendix shows calculations by various authors attempting to trace the possible losers from preference erosion. The Table shows that the countries with the relatively large losses (based on these studies) are Mauritius, Guyana and Swaziland. For these countries, the main losses are expected to arise from the loss of preferences for sugar under the ACP regime. Other countries such as Barbados and Fiji (sugar), and Botswana and Namibia (meat) also face losses according to some studies. Preference erosion has also been a theme related to the elimination of textile quotas, where some countries fear losing market shares to China and other strong suppliers after 1.1.2005 (see e.g. Mlachila and Yang 2004). Looking at competition with China as a zero-sum game may however not be appropriate: Lall and Albaladejo (2004) show that at least in East Asia, China is promoting rather than replacing the exports of its neighbours.

This however depends on the technology characteristics of exports, and exports of low-tech labour-intensive goods may have more problems with the direct competition from China. As noted above, further growth in China may also be important for other developing country in a long-run dynamic perspective.

Regarding losses from preference erosion, the method applied is important for the results. Some studies do not fully account for the increase in world market prices due to liberalisation. For example, the sugar market is heavily distorted by export subsidies and ACP countries losing some of their sales to EU could increase their sales to other countries. Norway is a good example; we import more all less all our sugar consumption from the EU, where it has been refined after the raw materials have been imported from developing countries. As shown by e.g. Milner et al. (2003), some ACP will gain rather than lose from EU sugar reform, if changed world market prices are accounted for. Using the world trade model GTAP, Kerkelä and Huan-Niemi (2005) also find that some countries in Sub-Saharan Africa gain from reform, and for some of them, the change is modest. A substantial loss for Mauritius nevertheless

20 In WTO document TN/MA/W/51/Add.1, dated 11 March 2005, The ACP countries suggest methodologies for identifying products for which preference erosion is important.

The implicit proposal is that MFN liberalisation should be slower for these products.

remains. And the world trade model clearly shows that Brazil is the big winner.

Estimates on preference erosion generally also assume that the rent from preferences is fully captured by the exporting country. In recent years, research on rent sharing has made it clear that this assumption is not plausible. The rent due to a tariff preference could be captured by the exporter, by the importer, by the importing country (if there are auctions), or by the consumers (if the price is lowered). Olarreaga and Özden (2005) examine prices for clothing exports to the U.S. under the preference scheme AGOA (the African Growth and Opportunity Act), and conclude that exporters on average receive 1/3 of the tariff rent. The poorer is the exporting country, the lower is this share. The more concentration there is on the importer side, the smaller is the rent obtained by the exporter. Silva (2005) analysed US-Latin America trade and found that importers capture an even higher share of the rent due to preferences, in some cases more than 95%.²¹ If these estimates should turn out to be representative, some estimates on the losses from preference erosion may be reduced substantially. Losses in the form of reduced employment and value added may remain, however.

A survey undertaken among a limited number of Norwegian importers (25) in the agriculture sector indicated that entry barriers are higher in the poorest countries, and the presence of limited competition in these markets is therefore more likely. In the light of the study referred to above, an interesting issue is whether the rents from exclusive preferences to the poorest countries may to a lesser extent be transferred to the exporters.

While we consider this to be likely, it need not be the case – if the exporters also have market power.

Regarding textiles and clothing and the fear of preference erosion, Norway is an interesting case since quotas have been abandoned before this occurred in the EU and the USA.²² In the case of Norway, Bangladesh benefited from the absence of quotas during the late 1980s and early 1990s, when other suppliers were restrained. This led to a considerable increase in Norway’s textile imports from Bangladesh, as shown in Diagram 5.

21 Silva also shows that lobbying from importers is an important determinant of the magnitude of trade preferences. Rents from preferential imports may be reason from this.

22 We have used the past form since this should have happened from 1 january 2005.

However, some new restrictions have been introduced later, so it may be too early to celebrate the complete elimination of textile quotas in the EU or the USA.

Diagram 5: Norway's imports of textiles and clothing from Bangladesh, 1976-2004 (Data source: COMTRADE)

0 10 20 30 40 50 60

1976 1978

1980 1982

1984 1986

1988 1990

1992 1994

1996 1998

2000 2002

2004

Mill. USD

Textiles (SITC 65) Clothing (SITC 84)

From the mid-1980s, other suppliers were restrained by MFA agreements while Bangladesh could expand freely. In addition, Bangladesh had zero tariffs as an LDC. The phasing-out of Norwegian textile quotas during the 1990s, and the reduction of Bangladesh’s tariff preference, has certainly led to a strong erosion of the Bangladesh preference. Clothing imports have nevertheless continued to accelerate, so here there is no sign of adverse effects.²³ Hence at least in the case of Norway, the fear of losses due to the elimination of textile quotas has not materialised.

On the whole, analysis of preference erosion as well as general equilibrium assessment of WTO liberalisation suggest that the problem of preference erosion is important for a limited number of countries concerned, but in pure economic terms of little importance for the world economy. The trade policy repercussions of preference erosion may however indirectly lead to greater economic effects: If preference erosion has the political effect of hindering trade liberalisation to the benefit of other developing countries, it will have a strong negative impact on others – in the form of “gains from trade liberalisation foregone” (see also Limão and Olarreaga 2005).²⁴ Özden and Reinhardt (2003) provide empirical support for the hypothesis that GSP tends to make beneficiary countries less liberal in trade policy. Hence the potentially serious costs of preference erosion for the world economy is more related to its indirect political than its direct economic effect. As also shown by Hoekman et al.

(2002), a uniform non-discriminatory tariff combined with direct

23 Bangladesh export data for SITC 84 Clothing are available from the COMTRADE database for 2001 and 2003. These suggest that there has not been a sharp import increase. We do not have an explanation of this. Given that Bangladesh ranks high on international lists of corruption, the possibility of re-routing of goods from other countries could be checked, or false declarations on the origin of goods. We do not, however, have any evidence suggesting that this is the case.

24 In order to avoid the political impact of preference erosion, the authors also suggest replacing preferences with an import subsidy for small developing countries.

assistance to the poorest countries may be better from a global welfare point of view.

The literature on preference erosion suggests that trade preferences certainly work, if they are large enough. On the other hand, temporary preferences that cannot be sustained in the long run may create serious adjustment problems in countries suffering from preference erosion. In this way, the help can turn into a problem. Trade preferences that are strongly discriminatory between different developing countries have adverse consequences for trade policy that may create losses for other developing countries by blocking trade policy progress.

As a consequence of these problems with preference erosion, it seems well advised to design GSP systems in order to avoid them in the future.

Mega-preferences of several hundred percentage points for a few countries should be avoided, in order to avoid non-unsustainable trade and subsequent adjustment problems. Preferences should be designed so that there is real competition between suppliers, in order to secure that the resulting trade is sustainable. This could either be obtained by including more countries, or by having “gradual differentiation” so that e.g. LDCs face competition from LICs or LMCs, LMCs face competition from UMCs and so on.

By delaying improvements in market access for other DCs, very large preferences to LDC, ACP and the like may on the whole have an adverse impact for developing countries. This is an additional argument for spreading the benefits more evenly. The “supply constraint” argument suggests that extended preferences should be given also to the “second poorest” in order to promote trade and welfare gains. While ways and means should be found to handle real problems of preference erosion, these problems should not be exaggerated – as our Bangladesh illustration suggests.

Growth in the more successful developing countries will also generate import demand that will promote exports from their neighbour developing countries. This has been shown in the case of China. Increased trade between adjacent developing countries has to be a core aspect of their development, and this should not be interrupted by “artificial” exports to Norway generated by mega-preferences. If increased trade between developing countries is slowed down due to the preference erosion dilemma, is will most likely have a negative overall impact on development.

9. Concluding remarks

The purpose of this paper has been to present a menu and some arguments relating to differentiation in GSP systems. The discussion is relevant for several issues with respect to Norway’s GSP system:

1. Should Russia and other Eastern European states obtain GSP?

According to income criteria, the answer would be affirmative.

2. Should Hong Kong and Korea be graduated from GSP? According to plausible definitions of developing countries, the answer is yes. In order to continue its free trade policy for these countries, Norway